Thursday, December 08, 2022
Improving WAR: Pitching
I recently asked on Twitter:
I will be improving WAR. Let me know what issues you may have.
I will then give you a score of 0 to 10 to tell you how much I share your concern.
Go...
I will go through the tremendous responses I received, and my concern level of 0 through 9. In this blog post, I'll deal with the item that received a concern level 10: Pitching WAR.
Background
Why is Pitching so difficult to figure out? It comes down to two things: Identification and Attribution.
It's easy enough to identify the pitcher when runs are scored and outs are recorded. It's another thing to attribute to the pitcher everything that happens when they are on the mound. Did the pitcher induce a double play, or did his fielders execute the double play? Did the pitcher get the strike that was an inch off the plate, or did the catcher frame that pitch? Did the pitcher allow a run when an outfielder's throw sails way over the catcher? Did the pitcher record an out when the outfielder nails the runner at home?
As you can see, lots of things happen. And sure, if we look at a pitcher's career, most of these things will cancel out, and what will be left standing is the pitcher themselves. But for a game? No. For a series of games? Mostly no. For a season? Sort of.
Current Two Paths
This is why the two most popular implementations of WAR has decided to approach it very specifically and from very different viewpoints: Baseball Reference starts with Runs and Outs (innings) and tries to adjust for all these non-pitcher impact points. Fangraphs starts with HR, BB, HBP, SO (aka FIP) and ignores the rest (on the idea that over the course of a season, most of those things will cancel out). So, both identify the pitcher, and then attribute to the pitcher those things I listed.
You will occasionally get disagreements, sometimes pretty wildly. Look no further than Cy Young winner Sandy Alcantara. On Baseball Reference, driven mostly by his very low 2.29 ERA, his WAR is whopping 8.0. On Fangraphs, driven mostly by his very good FIP of 2.99, his WAR is 5.7.
Which is right? Well, each are right for the path they took. They made a series of assumptions and each chose a single path to get there.
Three More Paths
What other paths could they have taken? Well, a third path is the wOBA path, where we focus on all the batter outcomes, but ignoring any baserunning events, as well as how those outcomes are sequenced (we don't care if hits are scattered or cluttered). Alcantara had a .260 wOBA, when the league average is .310. So his wOBA is 84% of league average, and if you square that, you get 70%, which is our estimate of his ERA relative to league average. Since the league ERA was 3.97, then 70% of that is 2.79. So, this third path is between the FIP path of 2.99 and the ERA path of 2.29, but leaning toward FIP (at least for Alcantara in 2022).
There's a fourth path, the xwOBA path, which is how well the batter hit the ball, but without concerning ourselves with the fielding alignment or fielder accomplishments. So, it's an offshoot of wOBA. From that standpoint, Alcantara had a .267 xwOBA, which at 86% of league average translates to a 2.92 xERA, very close to his FIP.
The fifth path is an offshoot of FIP: stick with his SO, BB, HBP, but instead of actual HR, use the "HR in X Parks" method (overlay all his actual batted balls across all 30 parks and count how many clear the fence). So instead of his actual 16 HR, we'd use his xHR of 18. That would increase his FIP by 0.11, or 3.10.
Summary and Next Step
So, there you go. Let's summarize:
2.29 ERA (runs and outs)
2.99 FIP (SO, BB, HBP, HR)
3.10 FIPx (SO, BB, HBP, xHR)
2.70 wOBA (outs, BB, HBP, 1B, 2B, 3B, HR)
2.92 xwOBA (PA, BB, HBP, xwOBAcon)
How much weight should we apply to each? I don't know yet. I do know that FIP is much better at attributing the actual accomplishments of a pitcher than ERA. My expectation is that each of the five paths will likely be similarly weighted. So, until I figure that out, I'd suggest just going with a simple average of all five. And therefore in this case, Alcantara will get an Attributed ERA of 2.80. So his Attributed WAR would likely come in at around 6.2
Recent comments
Older comments
Page 1 of 150 pages 1 2 3 > Last ›Complete Archive – By Category
Complete Archive – By Date
FORUM TOPICS
Jul 12 15:22 MarcelsApr 16 14:31 Pitch Count Estimators
Mar 12 16:30 Appendix to THE BOOK - THE GORY DETAILS
Jan 29 09:41 NFL Overtime Idea
Jan 22 14:48 Weighting Years for NFL Player Projections
Jan 21 09:18 positional runs in pythagenpat
Oct 20 15:57 DRS: FG vs. BB-Ref
Apr 12 09:43 What if baseball was like survivor? You are eliminated ...
Nov 24 09:57 Win Attribution to offense, pitching, and fielding at the game level (prototype method)
Jul 13 10:20 How to watch great past games without spoilers